Upload
eddy-blacker
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
Air Traffic Analysis, Inc
Using WITI for Airport Arrival Performance Analysis
A report on work-in-progress
December 2010
2
WITI and WITI-FA (“Forecast Accuracy”)
WITI = “Weather weighted by traffic”
• En-route weather• E-WITI uses actual convective Wx data, e.g. NCWD
• E-WITI-FA uses convective forecast data, e.g. CCFP, LAMP, …
• Both use the same scheduled traffic on major flows
• Convective forecast data is converted to “quasi-NCWD” format
• Terminal weather• T-WITI uses actual surface Wx data (METARs)
• T-WITI-FA uses surface Wx forecast data (TAFs)
• Both use the same scheduled traffic at major airports
• TAF converted to quasi-METAR form, “rolling look-ahead” stream
3
VFR
IFR
VFR
Weather Event
Arriv
al R
ates
Time
Airport Capacity Rate
Actual Arrivals
region of possible avoidable costs
Framework for Quantifying Avoidable and Unavoidable Weather Impact
4
Arrival Rate Deficit: Illustration
0
5
10
15
20
25
30
35
40
45
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
AAR
Hour, Z
LGA, Apr 3, 2009 Arrival Rates
Actual arrivals
Arr rate based on METAR
Arr rate based on 4hr TAF
Scheduled arrivals
Example: LGA, 04/03/2009
Forecast called for rain, low ceilings and strong winds from the southwest which would have forced LGA into a single-runway operation with low arrival rate.
Actual winds were much weaker. Ceilings lifted earlier than forecast.
Deficit between scheduled and actually-achieved arrival rates needs to be measured
Portion attributable to inaccurate weather forecast needs to be quantified
This portion then needs to be split into two pieces: see next slide
5
Accidents, outages, VIP flights, security, etc
0
5
10
15
20
25
30
35
40
45
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
AAR
Hour, Z
LGA, Apr 3, 2009 Arrival Rates
Actual arrivals
Arr rate based on METAR
Arr rate based on 4hr TAF
Scheduled arrivals
Avoidable
UnavoidableReasonable risk mgmt
Impact caused by other airports
Inaccurate forecast
Overly conservative TMI
Actual AAR < TMI (in excess of risk mgmt)
Deficit
Bad Weather
Avoidable and Unavoidable Portions of Arrival Rate Deficit: Breakdown
Unavoidable portion of arrival rate deficit (gap between “scheduled” and “maximum-achievable-given-actual-Wx” arrival rates) needs to be subtracted from overall deficit.
A reasonable risk mitigation factor (% arrival rate?) should also be subtracted.
Impact caused by other airports, unrelated to this one, should be subtracted as well.
What’s left is the avoidable portion. It can be subdivided into 4 categories: (a) Deficit caused by inaccurate forecast; (b) Overly conservative TMI – related to (a) and possibly partly caused by it; (c) Actual AAR below TMI’s (even after discounting for risk factor), and (d) Transition.
En-route to Terminal to Final Transition
6
Adding GDP InformationGDP and non-GDP Periods
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
Avoi
dabl
e D
elay
AAR
Hour, Z
EWR, Sep 30, 2010 Arrival Rates
Avoidable Delay
Actual arrivals
Arr rate based on Obs
Arr rate based on 4hr TAF
Scheduled arrivals
GDP Rate
Non-GDP
GDP
Non-GDPGDP
7
Non-GDP Arrival Rate Deficit Dissection
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
Avoi
dabl
e D
elay
AAR
Hour, Z
EWR, Sep 30, 2010 Arrival Rates
Avoidable Delay
Actual arrivals
Arr rate based on Obs
Arr rate based on 4hr TAF
Scheduled arrivals
GDP Rate
Not counted (scheduled arrival rate is too low)
Avoidable delay = 0 (actual arr rate >= scheduled)
Over-forecast and actual arr rate < scheduled: • all the deficit goes toward avoidable
delay (“Non-GDP Inefficiencies”);
• a portion of the deficit is attributed to Wx forecast inaccuracy
-100
0
100
200
300
400
500
600
700
800
More Onus on GDP Setup Less Onus on GDP Setup
Avoi
dabl
e D
elay
, Hou
rs
Avoidable Delay Estimates Including GDP Efficiency and GDP Execution, EWR, 09/30/2010
Non-GDP Inefficiencies
Non-GDP Wx Fcst Inaccuracy
GDP Wx Forecast Inaccuracy
GDP Setup Inefficiencies
GDP Execution Inefficiencies
Effect of Other Airports
GDP assessment includes a 7% risk management AAR credit(reduces avoidable delays)
Other airports' departure delaysare converted to arrival delaysfor affected destinations butonly for those hours when theiravoidable delays were > 0
Non-GDP I neff
Non-GDP Wx FcstInacc
Non-GDP
8
GDP Arrival Rate Dissection
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
Avoi
dabl
e D
elay
AAR
Hour, Z
EWR, Sep 30, 2010 Arrival Rates
Avoidable Delay
Actual arrivals
Arr rate based on Obs
Arr rate based on 4hr TAF
Scheduled arrivals
GDP Rate
No impact (GDP > scheduled rate)
GDP Setup Inefficiency (but avoidable delay due to GDP Wx forecast inaccuracy = 0)GDP Execution
Inefficiencies
-100
0
100
200
300
400
500
600
700
800
More Onus on GDP Setup Less Onus on GDP Setup
Avoi
dabl
e D
elay
, Hou
rs
Avoidable Delay Estimates Including GDP Efficiency and GDP Execution, EWR, 09/30/2010
Non-GDP Inefficiencies
Non-GDP Wx Fcst Inaccuracy
GDP Wx Forecast Inaccuracy
GDP Setup Inefficiencies
GDP Execution Inefficiencies
Effect of Other Airports
GDP assessment includes a 7% risk management AAR credit(reduces avoidable delays)
Other airports' departure delaysare converted to arrival delaysfor affected destinations butonly for those hours when theiravoidable delays were > 0
GDP Execution Ineff
GDP
9
Avoidable Delay/Cost analysis work is in progress and will continue in FY11
Another WITI application: airport delay prediction
(Both programs are funded by the FAA ATO-P Aviation Weather Group and led by AvMet Applications, Inc)
10
Training WITI Model on Airport DelayExample: ATL, 2007 (WITI based on actual Wx)
0
20000
40000
60000
80000
100000
120000
140000
160000
180000
200000
ATL ASPM Delay (Daily Total), minutes, and ATL Airport WITI, 2007
WITI
Delay
Once trained on historical data, the Airport WITI model can be used for delay prediction using forecast Wx and scheduled traffic for the day
11
Hourly Delay Forecast, 06/08/2008 Forecast: LAMP (convective); TAF (surface)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
Delay, min
Hour, Z
EWR Actual and Forecast Delay by hour, 06/08/2008
Actual ASPM Delay, min
4hr WITI-FA forecast "Delay", min
12
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5
Delay, min
Hour, Z
EWR Actual and Forecast Delay by hour, 01/28/2009
Actual ASPM Delay, min
4hr WITI-FA forecast "Delay", min
Hourly Delay Forecast, 01/28/2009 Forecast: TAF (surface)
13
Back-up Slides
14
WITI: Measuring Weather / Traffic Impact “The Hand the NAS Is Dealt Every Day”
Traffic Demand
National Airspace System (NAS)
ATM, Airline Response Strategies
Operational Outcomes
Local Airport Weather
En-route Convective Weather
The Weather Index (WITI) expresses severity of weather impact on the NAS, weighted by air transportation service demands
Capacity, Safety constraints
15
Queuing Delay Modeled by Wx Index Software, PHL, Dec 14, 2006
0
30
60
90
120
150
0600Z Dec 14 - 0600Z Dec 15, Hourly Wx Observations
Ho
urly
Dem
and
, Est
imat
ed C
apac
ity,
an
d D
elay
Dep+Arr demand
Dep+Arr capacity
Queuing Delay, Hrs
Actual queuing delays at PHL on 12/14/06 were significant (223 flights delayed > 15 min) and there were approx. 100 cancellations (so the resulting queuing delay was less than it could have been if there were no cancellations)
Optimum capacity in good Wx
WITI is a weighted sum of three components:
WITI CompositionWeather Weighted by Traffic, Quantified
– En-route Component: hourly frequency on major flows X amount of convective Wx that these flows cross
– Terminal Component
15
Used by the FAA and NWS on a regular basis:• Measure system performance in an objective manner – weekly reports • Compare different seasons’ Wx/traffic impact with outcomes (e.g. delays)
– Linear part: capacity degradation due to terminal weather impact, proportional to number of ops
– Non-linear (Queuing Delay) part reflecting excess traffic demand vs. capacity
16
Method: Use Airport Arrival RatesCompute Arrival Rate Deficit
We compare:
· Scheduled arrival rates from ASPM database
· Actual arrival rates, also from ASPM
· Model-generated arrival capacity based on METARs (i.e., actual weather data)
· Model-generated rates capacity based on TAFs (i.e., forecast weather data)
· Computed using a parametric model of airport capacity under different Wx conditions
· Use FAA’s airport capacity benchmarks and historical data on actual airport throughput
Any arrival rate deficit (“possible minus actual”) may be an indication of avoidable delays / cancellations
17
Re-Tooling WITI as an Airport Model
Standard WITI is a NAS Wx Impact assessment tool
A weighted sum of 3 components
– Weights computed to provide best correlation between WITI and Delay for OEP34 airports combined
WITI can be re-tooled as an airport model / delay predictor
– Use an airport specific, much more detailed WITI metric and “train” it on that airport’s delay-vs-Wx-and-traffic-demand data
– 12 components instead of 3 (“ATL Wind WITI, EWR Snow WITI, ORD Convective WITI”, etc)
Calibrate WITI straight to minutes-of-delay for direct comparison with actual ASPM delays